CN101833764A - Thymus section multi-scale image segmenting method - Google Patents
Thymus section multi-scale image segmenting method Download PDFInfo
- Publication number
- CN101833764A CN101833764A CN 201010159998 CN201010159998A CN101833764A CN 101833764 A CN101833764 A CN 101833764A CN 201010159998 CN201010159998 CN 201010159998 CN 201010159998 A CN201010159998 A CN 201010159998A CN 101833764 A CN101833764 A CN 101833764A
- Authority
- CN
- China
- Prior art keywords
- image
- scale image
- thymus
- subimage
- cluster
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The invention discloses a thymus section multi-scale image segmenting method, which comprises the following steps: firstly, linearly segmenting a primary image into a plurality of sub-image blocks and establishing a multi-scale image data relationship; secondly, acquiring information of neighboring sub-image blocks, performing downward acquisition processing to generate a clustering scale image, and performing image clustering processing to generate a clustering scale sub-image; and finally, performing upward acquisition processing to realize thymus section multi-scale image segmentation. The method of the invention solves the problem of high intensity caused by artificial analysis, also improves the precision of segmentation, can be used for identifying thymus tissues and provides base for diagnosis. When a thymus section tissue image is acquired, the calculated amount of further precise segmentation is reduced because the background noise is removed.
Description
Technical field
The present invention relates to be a kind of image processing techniques in the application on medical field method, specifically be a kind of thymus section multi-scale image segmenting method.
Background technology
Thymus gland (thymus) is the important lymphoid organ of body.Its function and immunity are closely related, secretion thymin and hormonal substance, the organ of tool endocrine dysfunction.Thymus gland is considered to play an important role in the morbidity to myasthenia gravis (MG).
Find by prior art documents, Wekerle, H. with M ü ller-Hermelink, H. (The thymus inmyasthenia gravis.Current topics in pathology, 1986,75:179), and Palace, J., Vincent, (Myasthenia gravis:diagnostic and management dilemmas.Current Opinion inNeurology, 2001 such as A., 14 (5): 583) at the pathologic finding of thymus gland to myasthenia gravis (myasthenia gravis, morbidity MG), the diagnosis and treat significant.Mainly rely on veteran pathology expert in microscopic field one by one paracytic quantity and position to be discerned for thymus section, the intensity of manual analysis is big, operating personnel's fatiguability, and personal error is inevitable.In addition, the thymus section data of large volume document formula are for the storage and the utilization also quite inconvenience of data.
Simultaneously, find by prior art documents, Ma Chunmei, Liu Guiru and Wang Lulin (applied research of image Segmentation Technology in Medical Image Processing. Taiyuan science and technology, 2007,28 (3): 64-67) and Ma, Z., Tavares, J. with Jorge R. (Segmentation of structures in medical images:review and a new computationalframework.In Proceedings of the CMBBE 2008-8th International Symposium on ComputerMethods in Biomechanics and Biomedical Engineering, 2008.) along with the image processing techniques develop rapidly, various image segmentation and sorting technique are also very extensive in application on medical field.So utilize digital image processing techniques, the thymus section image carried out multi-scale division.Automatically obtain the thymic tissue image, alleviated personnel's working strength.Reduced the calculated amount of further segmentation sectioning image tissue simultaneously.
Summary of the invention
The present invention is directed to the prior art above shortcomings, a kind of thymus section multi-scale image segmenting method is provided, solve personnel's fatigue and error problem that original identification brings, both solved the big problem of intensity that manual analysis causes, also improved the precision of tissue segmentation, can be used for the identification of thymic tissue, for diagnosis provides foundation.When obtaining the thymus section tissue image, owing to removed ground unrest, thus also reduced the calculated amount of further fine segmentation.
The present invention is achieved by the following technical solutions, the present invention includes following steps:
The first step is divided into the original image linearity experimental process image block and makes up the multi-scale image data relationship;
The area of described experimental process image block equates.
Described multi-scale image data relationship is meant: the subimage block on the top, the upper left corner of original image is designated as (1,1), according to European coordinate all subimage blocks are carried out two-dimensional coordinate and be labeled as (x, y), wherein x is the position of the row at subimage block place, and y is the position of the row at subimage block place.
Second step, obtain the adjacent sub-images block message, promptly each subimage block obtains its coordinate information of other adjacent subimage blocks on every side successively, and its coordinate information is recorded in this center subimage block;
In the 3rd step, the downward sampling processing of image generates cluster grade image: obtain cluster grade image after each subimage block is reduced the resolution sampling.
The 4th step, image clustering is handled and is generated cluster grade subimage: cluster grade image is synthetic again according to the multi-scale image data relationship of original foundation, after by clustering processing image being carried out front and back scape classification then, be divided into several cluster grade subimages according to linearity.
Described clustering processing is meant: adopt non-supervision K average algorithm cluster, promptly the K average algorithm is accepted input quantity k; Then n data object is divided into k cluster so that make the cluster that is obtained satisfy: the object similarity in the same cluster is higher; And the object similarity in the different clusters is less.The cluster similarity is to utilize the average of object in each cluster to obtain " center object " (center of attraction) to calculate.The course of work of K average algorithm: at first select k object as initial cluster center arbitrarily from n data object; And, then, respectively they are distributed to (cluster centre representative) cluster the most similar to it according to the similarity (distance) of they and these cluster centres for other object of be left; And then calculate the cluster centre (average of all objects in this cluster) of each new cluster that obtains; Constantly repeat this process till the canonical measure function begins convergence.Generally all adopt mean square deviation to have following characteristics as .k cluster of canonical measure function: each cluster itself is compact as much as possible, and separates as much as possible between each cluster.
The area of described several cluster grade subimages equates.
The 5th step, the image sampling processing that makes progress: the cluster subimage is improved the binary map subimage block of the thymic tissue that obtains being partitioned into after the resolution sampling, realize the thymus section multi-scale image segmentation.
The sampling of described raising resolution is meant: adopt in the 3rd step the inverse to the sampling ratio of down-sampling to carry out the binary map subimage block that resampling obtains.
The present invention solves personnel's fatigue and the error problem that original identification brings, and has both solved the big problem of intensity that manual analysis causes, and has also improved the precision of tissue segmentation, can be used for the identification of thymic tissue, for diagnosis provides foundation.When obtaining the thymus section tissue image, owing to removed ground unrest, thus also reduced the calculated amount of further fine segmentation.
Description of drawings
Fig. 1 is an embodiment of the invention multi-level images data relationship.
Fig. 2 is the original thymus section subimage block of the embodiment of the invention (part).
Fig. 3 is the thymus section subimage block (part) of embodiment of the invention process behind down-sampling.
Fig. 4 for the embodiment of the invention through image the thymus section image after synthetic.
Fig. 5 is the average sorted thymic tissue image of embodiment of the invention process K.
The contrast images of the binary map subimage block of the thymic tissue that Fig. 6 goes out for embodiment of the invention atom image block and cluster segmentation.
Embodiment
Below embodiments of the invention are elaborated, present embodiment is being to implement under the prerequisite with the technical solution of the present invention, provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
The cell section that present embodiment adopts derives from the thymic tissue section, and concrete whole implementation process may further comprise the steps as shown in Figure 1:
1, makes up the multi-scale image data relationship: original image grid linearity is divided into subimage.The standard of cutting apart is mainly considered the size of image and the processing power of computing machine, for the ease of calculating, adopts the integral multiple of original image as far as possible.Present embodiment embodiment adopts each subimage block 2000x2000 pixel to divide original image.Each subimage block and each subimage block under the different resolution yardstick have just had the data message that is in the original image ranks separately.The subimage block in the original image upper left corner is designated as (1,1), then other subimage block correspondingly be designated as (x, y).Wherein x is the position of the row at subimage block place, and y is the position of the row at subimage block place.
The total pixel of subimage block number=original image/(the capable pixel X of subimage block subimage block row pixel) (1)
2, the peripheral information of subimage block is duplicated.According to the relation between each subimage block of setting up, obtain single subimage block other subimage block information of 3x3 neighborhood on every side one by one, the position of promptly supposing this subimage block is (x, y), then obtain (x ± 1, y ± 1), the information of 8 field subimage blocks, and the image information of these neighborhoods copied in this subimage block, each subimage block just has the subimage information of expanding accordingly separately like this.The subimage block of expansion is carrying out just can not causing the error of resampling because original boundaries is discontinuous in down-sampling.
3, the subimage block that obtains is carried out to down-sampling.By resampling, adopt neighbour's interpolation (Nearest-neighborinterpolation), obtain the image of a low resolution.The downscaled images size in principle can be any, but mandatory foundation is to guarantee that the intensity profile of gray distribution of image and original image totally is consistent behind down-sampling.For the ease of calculating, will to the yardstick rounding of down-sampling the integral multiple of former figure size in addition.Present embodiment embodiment is with 400 times of original image minifications.As Fig. 2 result such as Fig. 3 behind down-sampling.
4, with the image behind the resampling, promptly cluster grade image is synthetic again according to the multi-scale image data relationship of original foundation, as Fig. 4.Because subimage is the subimage of the expansion after duplicating through the field, image after synthetic has again overcome the discontinuous problem in edge after image segmentation becomes subimage block.
5, the average clustering algorithm of K.The K average algorithm is accepted input quantity k; Then n data object is divided into k cluster so that make the cluster that is obtained satisfy: the object similarity in the same cluster is higher; And the object similarity in the different clusters is less.The cluster similarity is to utilize the average of object in each cluster to obtain " center object " (center of attraction) to calculate.The course of work of K average algorithm: at first select k object as initial cluster center arbitrarily from n data object; And, then, respectively they are distributed to (cluster centre representative) cluster the most similar to it according to the similarity (distance) of they and these cluster centres for other object of be left; And then calculate the cluster centre (average of all objects in this cluster) of each new cluster that obtains; Constantly repeat this process till the canonical measure function begins convergence.Generally all adopt mean square deviation as the canonical measure function.K cluster has following characteristics: each cluster itself is compact as much as possible, and separates as much as possible between each cluster.After the average clustering algorithm operation of K, the front and back scape of thymus section obtains good discrimination (as Fig. 5), and promptly whole thymic tissue is cut apart well.After removing the background noise, reduced the calculated amount of follow-up further fine segmentation.
6, according to cutting apart, obtain the cluster subimage block according to the image of the multi-scale image data relationship of setting up before after with cluster.
The subimage block number of cutting apart=
The total pixel of binary map after the cluster/(the capable pixel X of cluster subimage block cluster subimage block row pixel) (2)
7, image is synthetic, and the image after cutting apart according to the previous inverse of sampling rate downwards, is returned the binary map subimage block of the thymic tissue that is partitioned into after original size obtains cluster to up-sampling.As Fig. 6, be the contrast images of the two-value subimage block of thymic tissue after atom image block and the cluster segmentation.
The thymus section multi-scale image segmenting method of present embodiment utilizes digital image processing techniques to realize cutting apart of thymus section tissue.Utilize MATLAB, a kind of thymus section multi-scale image segmenting method is provided, make it solve personnel's fatigue and error problem that original identification brings, both solved the big problem of intensity that manual analysis causes, also improved the precision of tissue segmentation, can be used for the identification of thymic tissue, for diagnosis provides foundation.With one of present embodiment embodiment complete thymus section, size of data is the 542M byte, and view data is quite huge, and the computing machine that common sorting algorithm is applied to common arithmetic capability will cause internal memory to overflow.Adopt multi-scale image segmenting method, then significantly reduced the calculated amount of single step, overcome the problem that data are overflowed.By calculating, the sorted thymus section tissue image of acquisition accounts for 43.7% of original image, has promptly removed in the raw data 56.3% background noise information, thereby has reduced the calculated amount of further fine segmentation largely.
Claims (9)
1. a thymus section multi-scale image segmenting method is characterized in that, comprises the steps:
The first step is divided into the original image linearity experimental process image block and makes up the multi-scale image data relationship;
In second step, obtain the adjacent sub-images block message;
In the 3rd step, the downward sampling processing of image generates cluster grade image:
In the 4th step, image clustering is handled and is generated cluster grade subimage:
In the 5th step, the image sampling processing that makes progress realizes the thymus section multi-scale image segmentation.
2. thymus section multi-scale image segmenting method according to claim 1 is characterized in that, the area of the experimental process image block described in the first step equates.
3. thymus section multi-scale image segmenting method according to claim 1, it is characterized in that, multi-scale image data relationship described in the first step is meant: the subimage block on the top, the upper left corner of original image is designated as (1,1), according to European coordinate all subimage blocks are carried out two-dimensional coordinate and be labeled as (x, y), wherein x is the position of the row at subimage block place, and y is the position of the row at subimage block place.
4. thymus section multi-scale image segmenting method according to claim 1, it is characterized in that, described second step specifically is meant: each subimage block obtains its coordinate information of other adjacent subimage blocks on every side successively, and its coordinate information is recorded in this center subimage block.
5. thymus section multi-scale image segmenting method according to claim 1 is characterized in that, described the 3rd step specifically is meant: obtain cluster grade image after each subimage block is reduced the resolution sampling.
6. thymus section multi-scale image segmenting method according to claim 1, it is characterized in that, described the 4th step specifically is meant: cluster grade image is synthetic again according to the multi-scale image data relationship of original foundation, after by clustering processing image being carried out front and back scape classification then, be divided into several cluster grade subimages according to linearity.
7. thymus section multi-scale image segmenting method according to claim 1 is characterized in that, described the 5th step specifically is meant: the binary map subimage block that the cluster subimage is improved the thymic tissue that obtains being partitioned into after the resolution sampling.
8. according to claim 1 or 6 or 7 described thymus section multi-scale image segmenting methods, it is characterized in that the area of described several cluster grade subimages equates.
9. thymus section multi-scale image segmenting method according to claim 7 is characterized in that, the sampling of described raising resolution is meant: adopt in the 3rd step the inverse to the sampling ratio of down-sampling to carry out the binary map subimage block that resampling obtains.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010101599981A CN101833764B (en) | 2010-04-29 | 2010-04-29 | Thymus section multi-scale image segmenting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010101599981A CN101833764B (en) | 2010-04-29 | 2010-04-29 | Thymus section multi-scale image segmenting method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101833764A true CN101833764A (en) | 2010-09-15 |
CN101833764B CN101833764B (en) | 2012-04-25 |
Family
ID=42717825
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2010101599981A Expired - Fee Related CN101833764B (en) | 2010-04-29 | 2010-04-29 | Thymus section multi-scale image segmenting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101833764B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102564923A (en) * | 2011-12-29 | 2012-07-11 | 上海交通大学 | Device for automatically identifying fiber distribution density of bone marrow biopsy |
US9438769B1 (en) | 2015-07-23 | 2016-09-06 | Hewlett-Packard Development Company, L.P. | Preserving smooth-boundaried objects of an image |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106682419B (en) * | 2016-12-27 | 2019-05-07 | 深圳先进技术研究院 | Fitting method and device for medical image parameters |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002001143A2 (en) * | 2000-06-27 | 2002-01-03 | Electro-Optical Sciences, Inc. | Multispectral imaging and characterization of internal biological tissue |
CN101021944A (en) * | 2007-03-14 | 2007-08-22 | 哈尔滨工业大学 | Small wave function-based multi-scale micrograph division processing method |
WO2008008530A2 (en) * | 2006-07-14 | 2008-01-17 | The State Of Oregon Acting By And Through The State Board Of Higher Education On Behalf The University Of Oregon | Multiscale morphological topology correction of cortical surfaces |
CN101201937A (en) * | 2007-09-18 | 2008-06-18 | 上海医疗器械厂有限公司 | Digital image enhancement method and device based on wavelet restruction and decompose |
CN101526994A (en) * | 2009-04-03 | 2009-09-09 | 山东大学 | Fingerprint image segmentation method irrelevant to collecting device |
-
2010
- 2010-04-29 CN CN2010101599981A patent/CN101833764B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002001143A2 (en) * | 2000-06-27 | 2002-01-03 | Electro-Optical Sciences, Inc. | Multispectral imaging and characterization of internal biological tissue |
WO2008008530A2 (en) * | 2006-07-14 | 2008-01-17 | The State Of Oregon Acting By And Through The State Board Of Higher Education On Behalf The University Of Oregon | Multiscale morphological topology correction of cortical surfaces |
CN101021944A (en) * | 2007-03-14 | 2007-08-22 | 哈尔滨工业大学 | Small wave function-based multi-scale micrograph division processing method |
CN101201937A (en) * | 2007-09-18 | 2008-06-18 | 上海医疗器械厂有限公司 | Digital image enhancement method and device based on wavelet restruction and decompose |
CN101526994A (en) * | 2009-04-03 | 2009-09-09 | 山东大学 | Fingerprint image segmentation method irrelevant to collecting device |
Non-Patent Citations (2)
Title |
---|
《应用技术》 20070331 马春梅等 图像分割技术在医学图像处理中的应用研究 64-65、67 1-9 , 第3期 2 * |
《微计算机信息》 20100228 王申等 基于邻域的多尺度模糊C-均值聚类图像分割 184-185 1-9 第26卷, 第2期 2 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102564923A (en) * | 2011-12-29 | 2012-07-11 | 上海交通大学 | Device for automatically identifying fiber distribution density of bone marrow biopsy |
CN102564923B (en) * | 2011-12-29 | 2013-07-17 | 上海交通大学 | Device for automatically identifying fiber distribution density of bone marrow biopsy |
US9438769B1 (en) | 2015-07-23 | 2016-09-06 | Hewlett-Packard Development Company, L.P. | Preserving smooth-boundaried objects of an image |
Also Published As
Publication number | Publication date |
---|---|
CN101833764B (en) | 2012-04-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104881677B (en) | Method is determined for the optimum segmentation yardstick of remote sensing image ground mulching classification | |
CN101576997B (en) | Abdominal organ segmentation method based on secondary three-dimensional region growth | |
CN105303616B (en) | Embossment modeling method based on single photo | |
CN108447050A (en) | A kind of Surface Flaw dividing method based on super-pixel | |
CN108956420B (en) | Concrete pore feature extraction method and system | |
CN109087396A (en) | A kind of microscopical structure reconstructing method based on CT images of concrete pixel characteristic | |
CN103413286A (en) | United reestablishing method of high dynamic range and high-definition pictures based on learning | |
CN1912927A (en) | Semi-automatic partition method of lung CT image focus | |
CN110188778B (en) | Residential area element outline regularization method based on image extraction result | |
CN101587189B (en) | Texture elementary feature extraction method for synthetizing aperture radar images | |
CN111462838B (en) | Method for directly converting image pixels into finite element units | |
CN101833764B (en) | Thymus section multi-scale image segmenting method | |
CN109816664A (en) | A kind of three-dimensional point cloud dividing method and device | |
CN109345625A (en) | Core image adaptive partition three-dimensional rebuilding method | |
CN105550682B (en) | Bronze object stone inscription inscription rubbing method | |
CN112069704A (en) | Method for constructing multi-degree-of-freedom lattice model of real concrete specimen | |
CN109829896A (en) | The micro-calcification clusters automatic testing method of digital galactophore tomography X image based on multi-domain characteristics | |
CN111383207B (en) | Method, system and device for processing musculoskeletal ultrasound image and readable storage medium | |
CN111833432B (en) | Three-dimensional reconstruction method based on core two-dimensional gray scale image | |
CN110766609B (en) | Depth-of-field map super-resolution reconstruction method for ToF camera | |
CN109697693A (en) | A method of it realizes based on big data spatial operation | |
CN102437855B (en) | Data compression method of grade merging of octree three-dimensional model | |
CN111292339B (en) | Clinical temporal bone CT multi-structure extraction method and device based on W-type network structure | |
CN114581476A (en) | Lung organ model automatic leaf separation method and system based on CT image | |
CN112767548B (en) | Three-dimensional threshold value stereo graph unfolding method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120425 Termination date: 20180429 |